Research On Improving Real-time Performance Of BOTDA System

This dissertation analyzes the factors affecting the real-time performance of Brillouin Optical Time Domain Analysis(BOTDA) technology firstly,and summarizes the domestic and international research situations of dynamic BOTDA technology,Brillouin Frequency Shift(BFS) extraction method and BOTDA denoising method.In addition,the sensing principle of BOTDA and Rayleigh Brillouin Optical Time Domain Analysis(RBOTDA) and the structure of RBOTDA experimental system is introduced in detail.A new method of BOTDA signal denoising is proposed,which using the Total Variation(TV) model.The gradient descent is used for numerical solution,and the backtracking line search based on Armijo criterion is used to adjust the descent stepsize.The influence of Lagrangian factor on BFS extraction is analyzed,and the convergence speed using fixed stepsize and adaptive stepsize is compared each other.The RBOTDA temperature sensing experiment was carried out,and the Brillouin Scattering Spectrum(BGS) of 256 and 10,000 average was denoised according to the proposed algorithm.Then,the Lorentz Curve Fitting(LCF) was applied to extract BFS along the fiber.After removing the under-fitting data,the BFS of the heating and normal part of fiber were averaged and converted into temperature difference,and it matched the experimental condition.It is proved that TV image denoising can effectively remove system noise,ensure well BFS accuracy,and improve system real-time performance by reducing the average number of BOTDA.In addition,the computational complexity of LCF is analyzed and the basic principles of cross-correlation convolution and high-order centroid calculation are introduced in detail.A new method based on cross-correlation convolution and higher-order centroid algorithm for BFS extraction is proposed.The sensitivity of cross-correlation convolution to noise is simulated.The verification experiment of the proposed algorithm is carried out.The influence of point number and calculation order is explicated.The error performance of the proposed algorithm and LCF when using large sweep interval is compared.The experimental results show that it has similar curve shape with the BFS of LCF.It is feasible and doesn't have under-fitting phenomenon.It has lower computational complexity and higher reliability.Moreover,the appropriate number of points and order can restrict the error within 1MHz.If increasing the sweep stepsize to reduce measurement time,the error of the proposed algorithm is smaller than that of LCF.